25
Racial Disparities in Access to Mortgage Credit: Does Governance Matter? n Colleen Casey,University of Texas at Arlington Davita Silfen Glasberg, University of Connecticut Angie Beeman,CUNY-Borough of Manhattan Community College Objectives. This article examines the effect of community organizing on the like- lihood that minority borrowers pursue home mortgage credit from regulated lend- ers. Methods. Governance perspectives suggest that community organizations exert significant influence on policy outcomes. We use logistic regression with interaction terms to test the effect of community organizing on the lending outcomes of minority borrowers. We use a matched control sample of cities, drawing on 2004 loan data from two midwestern cities similar in racial and economic composition but with different histories of organizing around the Community Reinvestment Act (CRA). Results. We find differential effects based on an applicant’s race or eth- nicity. Overall, African-American applicants are less likely to pursue mortgage credit for home ownership from regulated lenders than their white, non-Hispanic coun- terparts. However, African Americans seeking mortgage credit in a city with a history of CRA organizing are more likely to apply to regulated lenders than their racial counterparts in a city without CRA organizing. However, while organizing reduces the disparities between white and African-American applicants, a gap still remains. Conclusion. African-American borrowers living in cities with a history of community organizing around CRA appear more likely to pursue mortgage credit from traditional, regulated lenders, suggesting that governance matters. Economic theories largely rooted in public choice theory have been ap- plied to analyze the impact of the Community Reinvestment Act (CRA) of 1977 (Barr, 2004), evaluate the ability of CRA to increase or decrease efficiency in lending markets (Benston, 1997; Lang and Nakamura, 1993), and its ability to increase access to credit from traditional lending institu- tions for historically disadvantaged borrowers. Although insight from these analyses provides relevant policy information, analyses rooted in these per- spectives adopt a traditional approach to policy analysis, one that assumes policies are implemented by a neutral administrative state in line with static, n Direct correspondence to Colleen Casey, Assistant Professor, School of Urban and Public Affairs, 526 University Hall, Box 19588, University of Texas at Arlington, Arlington, TX 76019-0588 o[email protected]. Colleen Casey will share all data and coding for replication purposes. We thank and acknowledge several anonymous reviewers and editors whose con- structive criticism was invaluable in making this a better article. Any errors or inconsistencies herein are entirely our own. SOCIAL SCIENCE QUARTERLY, Volume 92, Number 3, September 2011 r 2011 by the Southwestern Social Science Association DOI: 10.1111/j.1540-6237.2011.00792.x

Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

Embed Size (px)

Citation preview

Page 1: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

Racial Disparities in Access to MortgageCredit: Does Governance Matter?n

Colleen Casey,University of Texas at Arlington

Davita Silfen Glasberg, University of Connecticut

Angie Beeman,CUNY-Borough of Manhattan Community College

Objectives. This article examines the effect of community organizing on the like-lihood that minority borrowers pursue home mortgage credit from regulated lend-ers. Methods. Governance perspectives suggest that community organizations exertsignificant influence on policy outcomes. We use logistic regression with interactionterms to test the effect of community organizing on the lending outcomes ofminority borrowers. We use a matched control sample of cities, drawing on 2004loan data from two midwestern cities similar in racial and economic compositionbut with different histories of organizing around the Community Reinvestment Act(CRA). Results. We find differential effects based on an applicant’s race or eth-nicity. Overall, African-American applicants are less likely to pursue mortgage creditfor home ownership from regulated lenders than their white, non-Hispanic coun-terparts. However, African Americans seeking mortgage credit in a city with ahistory of CRA organizing are more likely to apply to regulated lenders than theirracial counterparts in a city without CRA organizing. However, while organizingreduces the disparities between white and African-American applicants, a gap stillremains. Conclusion. African-American borrowers living in cities with a history ofcommunity organizing around CRA appear more likely to pursue mortgage creditfrom traditional, regulated lenders, suggesting that governance matters.

Economic theories largely rooted in public choice theory have been ap-plied to analyze the impact of the Community Reinvestment Act (CRA) of1977 (Barr, 2004), evaluate the ability of CRA to increase or decreaseefficiency in lending markets (Benston, 1997; Lang and Nakamura, 1993),and its ability to increase access to credit from traditional lending institu-tions for historically disadvantaged borrowers. Although insight from theseanalyses provides relevant policy information, analyses rooted in these per-spectives adopt a traditional approach to policy analysis, one that assumespolicies are implemented by a neutral administrative state in line with static,

nDirect correspondence to Colleen Casey, Assistant Professor, School of Urban and PublicAffairs, 526 University Hall, Box 19588, University of Texas at Arlington, Arlington, TX76019-0588 [email protected]. Colleen Casey will share all data and coding for replicationpurposes. We thank and acknowledge several anonymous reviewers and editors whose con-structive criticism was invaluable in making this a better article. Any errors or inconsistenciesherein are entirely our own.

SOCIAL SCIENCE QUARTERLY, Volume 92, Number 3, September 2011r 2011 by the Southwestern Social Science AssociationDOI: 10.1111/j.1540-6237.2011.00792.x

Page 2: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

objective policy objectives (Salamon, 2002). Hence, they focus attention onthe relationship between formal regulatory bodies and the private entitiesthey regulate. Second, they focus on only one subset of lending institu-tions—those governed by formal CRA regulation.

Conversely, governance perspectives assume policies are rarely adminis-tered in a neutral state by formal governmental agencies (Salamon, 2002),but that policies, and their outcomes, are embedded in dynamic local socialand political contexts. As such, traditional approaches may neglect to con-sider the broader, institutional context in which policy implementation oc-curs, the different institutional arrangements through which policyobjectives are pursued, and the actors involved in governance in these con-texts. In this article, we focus on the case of community reinvestment andefforts aimed at stimulating access to credit for disadvantaged borrowers. Wetest if disadvantaged borrowers have a greater likelihood of pursuing accessto credit from traditional lending institutions in a context where local-levelactors have historically been engaged in community reinvestment efforts.We anticipate the likelihood of pursuing access to credit from traditionallenders will be greater for historically disadvantaged applicants in a placewith a history of community reinvestment efforts.

Perspectives of Governance

Because a host of formal and informal actors influence policy outcomes, abetter understanding of policy outcomes may be gained by adopting a gov-ernance perspective. As stated by Haines, Reichman, and Scott:

Given our understanding of the importance of informal governance withinpublic and private bureaucracies, and within societies more generally, whatcounts as ‘‘public’’ in some sense is no longer, if it ever was, the exclusivemonopoly of state bodies. . . . Rather it represents a more pluralistic way ofthinking about law and policy, where, in neither case, is the making orimplementation of policy exclusively within the control of governmentalorganizations. (2008:7)

Governance perspectives recognize the potential for a multitude of actorsand interests (although not necessarily equal in power) to influence policyoutcomes and action. Furthermore, governance perspectives challenge tra-ditional notions of a neutral administrative state consisting of governmentalagencies carrying out prescribed public policies, but instead suggest thatpolicy outcomes may be better understood by conceptualizing a blurredboundary between public and private actors (Stoker, 1998:17). For example,Schneider and Ingram (1997) argued that certain policy tools, such as ca-pacity tools, encourage greater participation among citizen or neighborhoodgroups by incorporating these informal actors in the process, in turn, in-fluencing the responses that emerge. Furthermore, Salamon (2002) argued

Racial Disparities in Mortgage Credit 783

Page 3: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

that policies contain elements that by design shift responsibility to a wholehost of actors that reside beyond the formal boundaries of government. Inshort, our understanding of policy outcomes and action may be betterinformed by understanding the ability of a wide range of groups and in-terests to shape the localized context.

Access to Credit: The Community Reinvestment Act of 1977

Arguably, an important and often-studied policy designed to stimulateaccess to credit in urban, minority communities is the 1977 CommunityReinvestment Act (CRA). One main goal of CRA is to stimulate reinvest-ment by increasing traditional lending activity in low-income, minoritycommunities. CRA was enacted in response to concerns that lenders wereredlining urban communities, that is, refusing to lend to borrowers based onthe characteristics of the community. Proponents of CRA argued that red-lining triggers a cycle of disinvestment in urban communities as both bor-rowers and businesses are unable to obtain access to credit (Squires, 1992,2003; Immergluck, 2004). The actual implementation and enforcement ofCRA has largely been characterized as involving not only formal, govern-mental actors, but also a whole host of informal actors, particularly com-munity-based groups.

Formal regulation under CRA is divided among the four federal agenciesresponsible for regulatory oversight in the financial industry and includes theFederal Reserve System (FRS), the Federal Deposit Insurance Corporation(FDIC), the Office of Comptroller of the Currency (OCC), and the Officeof Thrift Supervision (OTS). Regulators review the performance of thelender based on a preestablished exam schedule and can impose costs onlenders when their performance is deemed to be suboptimal. The regulatoryagency considers a bank’s CRA rating when the bank applies to open orclose a domestic branch, seeks to merge with another lending institution,consolidates, acquires assets, or assumes liabilities. If a bank has a less thansatisfactory CRA rating, a regulator could deny or delay a merger appli-cation, an acquisition request, or the opening of a new branch.

Formal agencies responsible for CRA oversight have interpreted CRA asholding lenders accountable in four areas: lending, investment, communitydevelopment, and service. Regulation under CRA, however, is not solelycarried out by federal regulatory agencies; lenders are often subject to chal-lenges posed by community-based groups or local governments. UnderCRA, community-based groups and other public entities are encouraged tochallenge or inform the behavior of lenders under CRA—efforts that maylead to a negotiated agreement. In a number of cities across the country,community-based and local groups have been extensively active in ensuringthat CRA is enforced: they have used it as a lever to negotiate agreementswith lenders for programs or policies to increase access to minority

784 Social Science Quarterly

Page 4: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

borrowers and to pressure regulatory agencies to review the lenders’ activityin certain communities (Immergluck, 2004; Joint Center for HousingStudies, 2002; Squires, 1992, 2003; Williams and Nesiba, 1997). In morerecent years, CRA agreements commonly occur when groups representingdifferent neighborhoods or communities, interested in economic justice, fairhousing issues, and/or community development, engage in collective actionto develop an active agenda to bring to the lenders.

Given the wide array of groups and interests beyond the formal bound-aries of government, the effects of CRA may best be understood from agovernance perspective. However, applying a governance perspective to un-derstand policy outcomes raises new questions about the relationship be-tween local actors, the interests they represent that shape the context inwhich local policies evolve, and the actions individuals take. Pluralistic the-ories suggest that public policy outcomes are realized largely through thepolitical process, whereby interests inform policy preferences, these interestsare representative of the racial and ethnic makeup of the residents in acommunity, public officials enact policy in response to these interests, andlocal governmental agencies neutrally administer policies through the use ofrules and standard operating procedures that create the desired conditions.Conversely, a governance perspective recognizes the potential for localities tobe shaped or characterized by different interests, the values espoused by theseinterests, and, in turn, how the involvement of these interests may shape notonly policy preferences and implementation, but also outcomes and behav-iors (Schneider and Ingram, 1997; Salamon, 2002). As such, governanceperspectives raise new questions about the different outcomes and oppor-tunities that might emerge in an environment governed by different interestsand perspectives.

A number of case studies suggest that this is indeed the case and thatlocalities are shaped by the interests and groups involved in governance(Bovaird, 2007; Hula and Jackson-Elmoore, 2001; Orr, 2001) and, as aresult, influence or shape the opportunities available to underserved or dis-enfranchised individuals. As summarized by Orr (2001), groups organizedaround certain issues or policy domains can change communities, improveneighborhoods, and expand opportunities at multiple stages of the policyand implementation processes. In doing so, they can inform and influencethe opportunities available to individuals who may have been previouslydisenfranchised from mainstream economic opportunities. In the arena ofenvironmental policy, for example, Gunningham, Kagan, and Thornton(2004) found that behavior in a community often reflects the businesses‘‘social license,’’ rather than the public policies in place, defined as thedemands on and expectations for a business enterprise that emerge fromneighborhoods, environmental groups, community members, and other el-ements of the surrounding civil society.

In this article, we focus on access to mortgage credit for homeownershipand the effect of localized community reinvestment interests. Researchers

Racial Disparities in Mortgage Credit 785

Page 5: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

continue to debate the extent to which CRA is actively enforced (Immerg-luck, 2004; Williams and Nesiba, 1997) and others suggest that CRA hasnever reached its full potential (Marsico, 2005), but there is a substantialevidence demonstrating CRA has stimulated reinvestment in urban com-munities (Bostic and Robinson, 2005; Immergluck, 2004; Schwartz, 1998,2000; Squires, 1992, 2003; Shlay, 1999; Williams, McConnell, and Nesiba,2001). Although the methods through which these studies reach these con-clusions differ, factors such as community-based organizing and agreements(Bostic and Robinson, 2005; Immergluck, 2004; Schwartz 1998, 2000;Squires 1992, 2003; Williams, McConnell, and Nesiba, 2001), internalcharacteristics of the formal regulatory agency such as culture or leadership(Immergluck, 2004), and presidential agendas (Immergluck, 2004) havebeen found to play a pivotal role.

Community-based groups’ engaged in CRA governance have largely beencharacterized as serving in a watchdog or regulatory capacity. In this ca-pacity, they influence both the behavior of regulatory agencies and lenders,which, in turn, has been found to influence the number of loans and dollaramount in CRA assessment areas. Most empirical analyses include variablesthat control for the number of CRA agreements that are present in an area,or the number of CRA agreements over time; focusing primarily on theformal effects of the policy such as its ability to correct market failures, anddevoting less attention to the groups present in a localized context. Othershave moved beyond explaining the effectiveness of CRA as merely a market-failure correction, and have explored the broader features of what has in-fluenced effective or strong CRA regulation, shedding light as to how thevariations in governance across place might influence the outcomes thatresult. For example, Williams, McConnell, and Nesiba (2001) found that ina county identified as having a high-level of activism that led to a negotiatedagreement, lenders were making a significantly higher percentage of loans inunderserved markets than in other markets. In this particular case, the in-volvement of nongovernmental actors influenced the behavior of lenders,leading to potential advantages for borrowers in this particular county.

The Broader Institutional Environment and Access to Credit

Studies such as the one above have called attention to the institutionalenvironments in which CRA-regulated lenders operate, yet much remains tobe known as to how the presence of these groups and interests influenceaccess to credit in a changed financial services environment. For Dymski(1999), CRA is just one piece of the puzzle of equal opportunity to capitaland credit markets, and merely one mechanism through which opportunitymight be shaped. The landscape of opportunities available to potentialborrowers has changed, and this environment may negate the original intentof CRA to encourage lenders to engage in outreach and identify oppor-

786 Social Science Quarterly

Page 6: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

tunities to provide access to mainstream financial services. The industry haschanged as a direct result of federal bank deregulation policy from onedominated by conventional mainstream lenders, to one characterized bydifferent types of lenders, from traditional savings institutions to mortgagebrokers and created financial service providers that are not subject to CRAregulation (Temkin, Johnson, and Levy, 2002), many of which specialize inoffering subprime credit. Lenders not regulated by CRA originate a higherpercentage of loans and higher percentage of subprime loans in minoritycommunities (Courchane, Surette, and Zorn, 2004; U.S. Department ofHousing and Urban Development, 2000). These new sources of credit,ungoverned by CRA legislation, may, in effect, merely repackage previousdisparities disguised as access to credit into new disparities in the form ofaccess to only certain types of lenders or institutions. Thus, a more completeunderstanding of access to credit in historically underserved communitiesextends beyond a discussion of CRA regulation per se to one that alsoconsiders the institutional structure of lending markets.

Likewise, it is also equally important to understand how social structureinfluences opportunities to access credit. Dymski and Aldana (2007) arguedthat ongoing racial inequality may be dictated by the social structure of aplace—namely, a particular group’s market power or share in a place. Intheir analysis of the Home Mortgage Disclosure Act (HMDA) loan data for84 metropolitan areas, they found the degree to which racial minorities facebarriers in accessing credit is related to their share of the market and theextent to which residential segregation is present in the areas where lendingtakes place: white applicants have the lowest odds of being approved for aloan in cities classified as ‘‘minority cities,’’ suggesting that the effects of aparticular policy cannot be considered without taking into account thebroader structural characteristics of the marketplace.

Taken together, these studies suggest a need to understand communityreinvestment outcomes as part of a larger lending environment, shaped andconstructed by community reinvestment interests, social structure, andpublic policies. However, while Dymski and Aldana (2007) illuminate theimportance of considering how the structural features of society might in-directly govern community reinvestment decisions, additional questions re-main. In their approach, power is conceptualized as something that resultsfrom population size or distribution in a place, but does not consider theeffects of different civic, community, and service networks in these envi-ronments—the mechanisms that connect, serve, and advocate the interestsof underserved borrowers and communities and may influence the oppor-tunities available to various racial and ethnic groups.

In this aricle, we extend Dymski and Aldana’s (2007) work and ask if thesimple number of minority residents in a community influences lendingdynamics, or is the mobilization or presence of groups representing theinterests of minority residents, those engaged in governance, an influencingfactor? For example, if we compare two cities with similar population dis-

Racial Disparities in Mortgage Credit 787

Page 7: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

tributions, do borrowers in these communities have the same likelihood ofaccessing traditional lenders as their counterparts in a city without a historyof groups focused on community reinvestment? If this matters, then onemight expect different outcomes among racial or ethnic minority borrowersin cities with similar populations of racial or ethnic groups but differentialhistories of organizing around community reinvestment issues. We begin toexplore this relationship by comparing two cities with different histories oforganizing around community reinvestment issues.

Disparities in Localized Interests: CRA Activity in Cleveland and St. Louis

The disparities in actors’ involved in community reinvestment efforts havebeen addressed in other literature. Holyoke (2004) and Casey (2009) un-cover the disparities across cities in the interests engaged in communityreinvestment issues. Community reinvestment efforts vary widely acrossstates, regions, and cities (Squires, 2003; NCRC, 2005). In some cities,groups have used CRA as a lever to negotiate agreements with lenders. Inother cases, groups have entered into partnership arrangements with lendersto deliver access to credit. In other cities, there is little or no evidence ofadditional action through CRA (NCRC, 2005, 2007), since its inception.

For example, Cleveland and St. Louis, two cities similar in regard toeconomic trends and socioeconomic and demographic characteristics ofresidents, have very different histories of organizing around communityreinvestment and access to credit. Cleveland has a long history of localorganizing around community reinvestment. Since the enactment of CRA in1977, local groups and public-sector officials in Cleveland have negotiated27 CRA agreements, and continued to negotiate agreements well into themid-2000s. Early CRA efforts were best characterized as reinvestment effortsby local neighborhood groups, but later transitioned to city-led reinvestmentinitiatives. Beginning with Mayor White’s administration, from 1989–2001,local government took the primary lead in negotiating lending agreements,based on an agenda developed in conjunction with a network of communitydevelopment corporations, efforts that continued through 2004. The use ofCRA has generated informal agreements between community-based groupsand lenders. These have stimulated linkages between city officials, commu-nity development corporation (CDC) networks, and lenders to developprograms that make mortgage loans for new homes available at low interestrates, include provisions for homeownership counseling, opportunities forwealth creation by combining mortgage loans with matched savings ac-counts, relaxed underwriting standards, ongoing lender involvement in as-sessing community credit needs, and lender commitments to matchedmarket share lending in low-income and minority communities (Bright,2003; National Community Reinvestment Coalition, 2005; Schwartz,1998, 2000).

788 Social Science Quarterly

Page 8: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

Conversely, neither community-based groups nor local government agen-cies in St. Louis have historically been actively engaged in CRA efforts. In St.Louis, during the years 1977–2004, community-based groups and organi-zations entered into only four CRA agreements, totaling $72.6 million(NCRC, 2005). All the agreements were negotiated during the 1990s, dur-ing the high time of merger activity among banks.

The efforts in Cleveland vary significantly from those in St. Louis. Com-munity activism and organizing around CRA in Cleveland led to a widearray of programs and agreements as well as encouraged the City of Cleve-land to engage in CRA activity as well. This makes Cleveland relativelyunique, as in very few cities have organizing efforts also influenced theinvolvement of local city government. This might suggest that communityactivism may be a significant factor in reinvestment success in cities, par-ticularly as it relates to reducing disparities in access to credit for members ofracial and ethnic groups.

Disparities in Access to Credit

Research shows that minority applicants are denied home mortgage creditat higher rates than white applicants with the same income and debt-to-income ratios (Dymski, 1999; Flippen, 2001). In particular, African Amer-icans are disadvantaged in all markets and by all lender types: on averageAfrican Americans are 36.3 percent less likely to receive access to credit,Hispanics 18 percent less likely, and Native Americans 36.3 percent lesslikely to receive such access, than their white counterparts. Conversely, AsianAmericans are more likely to receive access to credit, on average, and are 5.8percent more likely to be approved for a loan. Borrowers with lower incomesare at a disadvantage in obtaining home loans relative to applicants withhigher incomes, even controlling for other factors, as are applicants who seekto purchase homes in lower-income neighborhoods or in neighborhoodswith high proportion of minority residents (Dymski, 1999). Yet, there isgreat variation in how lenders treat minority borrowers and communitiesacross places: Dymski (1999) found no type of lender consistently serves allminority borrowers or neighborhoods better than any other type, and thisvaries within the same metropolitan area even among the same type oflenders. In some cases, lenders will serve one minority group better thananother.

Furthermore, since the mid-1990s, the issue facing many urban, low-income communities is the extent to which subprime institutions, typicallynot subject to direct CRA regulation, and lending are concentrated amonglower-income and minority borrowers. Subprime lending and institutionsspecializing in subprime credit have a strong correlation with borrower-specific characteristics, such as minority status and education levels (Courc-hane, Surette, and Zorn, 2004), as well as neighborhood characteristics such

Racial Disparities in Mortgage Credit 789

Page 9: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

as low-income, predominantly minority, and urban communities (Courc-hane, Surette, and Zorn, 2004; U.S. Department of Housing and UrbanDevelopment, 2000). Likewise, there is increasing evidence of a correlationbetween credit from nontraditional lenders and high rates of foreclosures inminority communities (Bunce et al., 2001; Gruenstein and Herbert, 2000).Observers debate whether these new lenders provide opportunities for bor-rowers unable to access credit through traditional lenders (Nichols, Pen-nington-Cross, and Yezer, 2004), or merely provide new opportunities forracial and class exploitation (Williams, Nesiba, and McConnell, 2005; Wylyet al., 2006). Given the original intent of CRA, it becomes necessary toconsider the disparities that exist among white and minority borrowers, notjust in access to credit, but also the types of lenders through which access isdifferentially governed. From the standpoint of equity concerns, more ap-propriate questions emerge, such as at what price is credit received and,perhaps more importantly, to what extent do disadvantaged applicants dis-proportionately seek access to credit from lenders not governed by CRAlegislation.

The Ongoing Fight for Access to Credit

Today’s lending environment is characterized by differences in the formalgovernance of lending institutions, regulated versus nonregulated lenders,and differences in the informal actors engaged in governance. The historicalcontexts in which policies and lending behavior emerge is central to theargument made in this article—that these contexts may influence differentialoutcomes for historically underserved groups, that these efforts extend be-yond one mere policy, and that these efforts may be more powerful thansocial structure alone. Even as the mortgage environment changed, and anew subset of lenders emerged that were not governed by CRA, reinvest-ment interests in Cleveland continued to push for additional actions to alterthe outcomes for underserved borrowers, whereas efforts in St. Louis re-mained nonexistent.

For example, emerging concerns over the prevalence of higher-pricedsubprime credit in minority communities prompted a refocus of communityreinvestment agendas in Cleveland. In 2002, the city enacted an ordinancedesigned to deter high-priced lending. However, in 2003, the city’s actionwas overruled by the state, as the Cuyahoga County Court of Appeals andthe Montgomery County Court ruled the ordinance invalid. Despite theruling, the city ordered the ordinance to remain in effect, until it was finallyoverruled in 2006. The three main objectives of Cleveland’s local ordinancewere to collect and make data available to the public to protect themselvesagainst predatory lenders, to prevent lenders engaged in predatory, or high-priced, lending from doing business with the city, and to prevent predatorylenders and those providing referrals and services to those lenders from

790 Social Science Quarterly

Page 10: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

taking advantage of city programs designed to encourage home ownershipand home repair. Regardless of the federal policy environment, local groupsrepresenting the interests of minority communities and borrowers continuedto push and alter the rules of the lending game.

However, while these concerns were prevalent in some cities across thenation, not all communities took additional action. For example, in St.Louis, no local policies or ordinances were enacted to protect residents frompredatory or high-cost lending. The significance of considering the contextin which community reinvestment takes shape is that the interests present inthese communities may shape the opportunities available to potential bor-rowers in these environments.

Institutional Space of Mortgage Credit: Methodological Approach

We build on the work of Dymski and Aldana (2007) by considering notonly the impact of social structure in a place on lending outcomes, but alsothe effects of local mobilization around community reinvestment. However,we extend their work in two ways. First, our sample differs because weanalyze data from two cities that are comparable in regard to minoritypopulation, and broader economic and regional characteristics, to isolate theeffects of historical organizing around CRA. We anticipate CRA activity willinfluence the opportunities available to minority groups.

Second, while Dymski and Aldana (2007) considered the effects of raceand place on access to mortgage credit, we add to their model by consideringthe changed opportunity structures presented in the current financial ser-vices environment in which lenders operate, one consisting of both CRA-regulated and nonregulated lenders, to account for the type of lender fromwhich an applicant pursues access to credit. Specifically, in this article, wefocus on the type of lender to which a potential borrower applies for credit,rather than the type of loan he or she receives or whether the potentialborrower accepted or rejected. We do this for several reasons. First, it isimportant to understand the types of lending institutions to which differentgroups have access or pursue access for homeownership. Equity concernsremain considering the extent to which traditional, mainstream lendersprovide opportunities to members of racial and ethnic groups. Second, theinformation and data available on the cost of loans—prime or subprime—isrelatively new, and a complete comparison of the actual type of loan receivedoften depends on a number of private factors that are not easily available.Finally, the focus of this analysis is on opportunity, and the extent to whichdisadvantaged borrowers have access to the same types of institutions as theircounterparts is a key focus of community reinvestment efforts and earlyCRA legislation. Thus, the central question in this article: Are historicallydisadvantaged applicants less likely to seek access from CRA-regulatedlenders? We anticipate they will be less likely to do so; however, these effects

Racial Disparities in Mortgage Credit 791

Page 11: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

will be mitigated by the history of actors involved in community reinvest-ment efforts in a place.

Logistic regression is used to test the effects of CRA activity. We estimateda logistic regression equation with interaction effects to test the effects of anapplicant’s racial or ethnic affiliation on type of lender from which a loanwas sought. The interaction represents the joint effect, in this particular case,of being a minority applicant in a city that has extensively organized aroundCRA. The cross-sectional sample draws on 2004 HMDA data from twocities, St. Louis, Missouri, and Cleveland, Ohio, and is restricted to ap-plications for single-family home loans. Data on home mortgage applica-tions were obtained from the HMDA database, which is the mostcomprehensive public database on mortgage loans and lenders; community-level characteristics were obtained from the Census 2000. To control forneighborhood and area effects, additional data were collected from theCensus 2000 Summary File 3 (SF3), available from the Census Bureau.

Whereas Dymski and Aldana (2007) control for regional differences inthe racial composition of these cities in their sample, we purposively selectthese two cities based on their similarity in racial and economic compo-sition. The distinguishing difference between the two cities, as noted pre-viously, is that Cleveland can be classified as a place with high communityreinvestment activity, whereas St. Louis can be classified as a place with lowcommunity reinvestment activity.1 As the analysis only focuses on two cities,it is plausible to argue that by focusing on only two cities, there is a potentialfor the results to be due to a spurious correlation. However, given theunique focus of our question concerning the localized patterns of commu-nity reinvestment across places, and how these efforts differ across differentcities, we felt a matched control case, that is, a city with little to no or-ganizing but exhibiting similar socioeconomic and demographic character-istics, was the preferred approach (Shadish, Cook, and Campbell, 2002).Given the assumptions on which our argument and analysis is based, wewere reluctant to add additional cities as we feel that it is unlikely that thegroups engaged in CRA activity in Cleveland are composed of the sameinterests and groups as one might find in CRA efforts in another city. Futurestudies that design matched-control cases, based on the interests and char-acteristics of the city, will enhance the ability to generalize causality.

No significant differences exist among the majority of racial and ethnicgroups in the sample cities, with the exception of borrowers identifying asAmerican Indian, Pacific Islander, Hawaiian, Native American, or Alaskan

1It is possible to argue that the reason Cleveland has a higher level of CRA activity isperhaps because it has more subprime or predatory practices in place that have influenced thisactivity. One factor that has been shown to influence CRA activity is bank merger activity,and these two cities experienced similar levels of merger activity, yet CRA activity is quitedifferent in these two cities (Casey, 2009). A historical account of reinvestment activity inCleveland suggests efforts toward reinvestment have been in place since 1979, and theseefforts continue to evolve as the lending environment changes (NCRC, 2005).

792 Social Science Quarterly

Page 12: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

Native. In 2003, the annual average unemployment rate for the cities ofCleveland and St. Louis was not significantly different, 8.1 and 8.5, re-spectively, and the regional rates did not vary significantly from each other,at 6.0 and 5.8, respectively (Bureau of Labor Statistics, 2008). Likewise, therates of homeownership at both the central-city level and the metropolitanlevel were not significantly different. Based on the SF3 data, the home-ownership rate in the City of Cleveland was 51.4 percent compared to a rateof 52.5 percent in the City of St. Louis. For both cities, the metropolitanrate of homeownership was much higher, but not significantly differentfrom each other, 72.4 percent for Cleveland and 73.3 percent for St. Louis.

Furthermore, St. Louis and Cleveland are also cities that are located instates that have similar regulations concerning intrastate branching, interstatebanking, and interstate branching, factors that influence the institutionallending environment and may influence the behavior of lenders. St. Louis didhave a slightly larger percentage of applications made to regulated lenders andCleveland had a slightly larger number of applications made to nonregulatedlenders. Of the 4,220 loan applications made for properties in Cleveland,51.5 percent were made to non-CRA-regulated lenders. Loan applications forproperties in St. Louis totaled 4,525, of which 45.3 percent were made tonon-CRA-regulated lenders. Lenders in Cleveland had a slightly higherorigination rate, 72.6 percent versus 70.3 in St. Louis. However, the percentof originations by regulated lenders was roughly the same: 82 percent in St.Louis compared to 81 percent in Cleveland. We anticipate in Cleveland, dueto historical activity around CRA, even given the slightly higher rate ofapplications made to nonregulated lenders, that minority groups that havehistorically stood at a disadvantage in lending markets will be more likely toapply to a regulated lender than will minority applicants in St. Louis.

Variables

The dependent variable, type of lender, is a binary variable with the valueof 1 (CRA-regulated lender) or 0 (non-CRA-regulated lender). CRA-reg-ulated lenders accounted for 52 percent of the sample. For the purposes ofthis study, a CRA-regulated lender is defined as a depository institutionregulated by one of the four federal agencies, the FDIC, the FRB, the OCC,or the Office of Thrift Supervision (OTS), and subject to CRA. An alter-native lender, or non-CRA-regulated lender, is a defined as a nondepositoryinstitution regulated by HUD. Nondepository institutions are for-profitlending institutions other than banks, savings associations, and creditunions. Nondepository institutions are subject to some regulation, includingthe HMDA; however, alternative lenders are not directly subject to CRA.2

2Applications to credit unions accounted for less than 1 percent of the sample and weredropped from the analysis.

Racial Disparities in Mortgage Credit 793

Page 13: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

The key independent variables include the racial or ethnic identificationof the loan applicant and whether the application was made in a city withCRA activity. An applicant’s racial or ethnic identification was determinedbased on the categories captured by HMDA data for primary applicants. Alimitation in this study is the categorization of an applicant’s racial andethnic identification. HMDA allows applicants to select among six racialcategories (American Indian or Alaskan Native, Pacific Islander or Hawai-ian, Native American, Asian, white, or black or African American) and oneethnic division (Hispanic or not Hispanic). Racial and ethnic categories werecollapsed into mutually exclusive categories, separating any applicants whoindicated Hispanic ethnicity into the Hispanic category. Because of thesmall number of applicants in each of the categories of American Indian orAlaskan Native, Pacific Islander or Hawaiian, and Native American, thesecategories were collapsed into one group. This resulted in applicants fallinginto one of five categories: American Indian, Alaskan Native, Pacific Is-lander, Hawaiian, or Native American; Asian; black or African American;white; or Hispanic. Loan applications where race was not reported weredropped from the analysis.3 To conduct the regression, the white categorywas dropped from the analysis to serve as the base group. The majority ofthe applicants in the sample were white, 54.8 percent. African-Americanapplicants constituted the second largest group of applicants in the sample(37.9 percent), followed by applicants identifying as Hispanic (4.9 percent),Asians (slightly less than 2 percent), and applicants falling into the broadcategory of American Indian, Alaskan Native, Pacific Islander, Hawaiian, orNative American (slightly less than 1 percent).

CRA activity is a binary variable, valued at 1 if the application was madein a city with a history of CRA activity (48 percent), and 0 if the applicationwas made in a city with little to no history of CRA activity (52 percent). Weconducted literature reviews and relied on a number of secondary publi-cations to determine if the city had a history of successful CRA activity(Casey, 2009; NCRC, 2005).

The sex (M 5 0.45, SD 5 0.498, 1 5 male) and income (M 5 48.42,SD 5 34.27, range of 10–432, in 1,000s) of the primary applicant were alsoincluded to control for any effects these may have on the type of lender towhich a potential borrower applies. Primary applicant data were consideredin the model because more information is captured on the primary appli-cant, that is, income, and because of the large amount of missing race and

3Sixty-four percent of applications of unknown race or ethnicity were made to nonreg-ulated lenders. A means comparison between individual and Census tract characteristics ofapplicants of unknown race or ethnicity (UN) and two subsets of the sample, nonwhite orHispanic applicants (NW) and white, non-Hispanic (W), did not suggest any observable biasin one direction or another. The means of each subset were compared on the characteristics ofapplicant income (NW: $40,760; UN: $47,540; W: $54,620), percentage of college ed-ucated in the Census tract (NW: 15.1; UN: 18.3; W: 25), tract income ratio (NW: 59.03;UN: 63.2; W: 76.8), and percentage of minority residents in the tract (NW: 70.3; UN: 57.2;W: 31.3).

794 Social Science Quarterly

Page 14: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

ethnicity data. Due to concerns of multicollinearity, loan amount wasdropped from the analysis. A series of other Census-tract-level variables thathave been found to have a significant relationship with alternative or non-regulated lenders were included in the model. The control variables includedthe following: percentage of minority population in the Census tract wherethe property was located (M 5 49.28, SD 5 24.84, range of 1.84–99.77),percentage of college educated residents in the Census tract (M 5 20.43,SD 5 12.52, range of less than 1–85.25 percent), total population in theCensus tract (M 5 3.56, SD 5 1.51, range of 0.074–9.154, in 1,000s), andnumber of housing units in the tract (M 5 1,710.66, SD 5 801.77, range of106–4,925, in 1,000s).4 Finally, the variable tract income ratio (M 5 68.70,SD 5 23.16, range of 14.64–143.65) captured the wealth of the Censustract, and wealthier tracts had a higher value.

Results

This section presents the results from the empirical analysis. Of mainconcern are the effects of an applicant’s race or ethnicity and CRA activityon the probability of applying to a regulated lender for a mortgage loan. Weanticipate that CRA activity will increase the probability that minoritygroups apply to regulated lenders, even when controlling for applicantcharacteristics. We estimate four logistic models in STATA version 11, onewithout interaction effects and no control variables (Model 1), a modelwithout interaction effects but including control variables (Model 2), amodel with interaction effects and no control variables (Model 3), and,finally, a model with interaction effects and control variables (Model 4).5

Based on model summary and fit statistics, we determine that Model 4 is thepreferred model. Table 1 reports the untransformed coefficients (coef),standard errors (s.e.), and probabilities (p) of the key variables of interest forall model specifications. Table 2 reports the transformation of the parameterestimates to odds ratios and as the percent change in odds ratio. The pre-dicted probabilities of the interaction terms are reported in Table 3. Finally,Table 4 reports the effects of the control variables.

4The STATA linktest indicated that the variables minority population, number of housingunits, and percent college educated residents exhibited nonlinear characteristics. After thevariables were transformed, STATA linktest and fittest suggested that the transformed vari-ables improved model fit.

5A number of studies have urged caution when interpreting the magnitude and signif-icance of interaction effects in logistic models, as heteroskedasticity can be a problem due tothe potential violation of the parallel lines assumption (Long and Freese, 2006; Norton,Wang, and Ai, 2004; Williams, 2009). There is an ongoing debate as to potential solutions toheteroskedasticity (for a full discussion, see Williams, 2009). One approach is to estimatemodels that relax the parallel lines assumption. In this article, gologit2 with the autofit optionwas used to relax the parallel lines assumption for the cases where it was justified; however,when examining the pattern of coefficients and comparing to Model 4, no distortions werefound.

Racial Disparities in Mortgage Credit 795

Page 15: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

TA

BLE

1

Untr

ansfo

rmed

Eff

ects

of

Race,

Eth

nic

ity,

and

CR

AA

ctivity

on

Apply

ing

toa

Regula

ted

Lender

Racia

land

Eth

nic

Chara

cte

ristic

sof

Ap

plic

ants

Mod

el1

Mod

el2

Mod

el3

Mod

el4

Coef

s.e.

pC

oef

s.e.

pC

oef

s.e.

pC

oef.

s.e.

p

Am

er.

Ind

.,A

lask

an,

Pac.

Isl.,

Haw

aiia

n,

or

Nativ

eA

mer.

�0.7

10n

0.2

53

0.0

05�

0.6

13n

0.2

57

.017

�0.4

00

0.4

23

0.3

44

�0.4

49

0.4

29

0.2

96

Asi

an

0.7

02n

0.1

86

0.0

00

0.7

77n

0.1

89

.000

0.4

70n

0.2

17

0.0

30

0.5

91n

0.2

22

0.0

08

Afr

ican

Am

eric

an

�1.0

52n

0.0

47

0.0

00�

0.5

28n

0.0

61

.000

�1.4

70

n0.0

69

0.0

00�

0.9

26

n0.0

81

0.0

00

His

panic

0.1

80

0.1

07

0.0

91

0.4

27n

0.1

10

.000

�0.1

61

0.1

97

0.4

15

0.0

80

0.2

02

0.6

92

CR

Aactiv

ity�

0.1

33n

0.0

45

0.0

03�

0.0

24

0.0

53

.652

�0.4

73

n0.0

60

0.0

00�

0.3

57

n0.0

68

0.0

00

Inte

ractio

ns:

CR

AA

ctiv

ityand

Race/E

thnic

ity

Am

er.

Ind

.,A

lask

an,

Pac.

Isl.,

Haw

aiia

n,

or

Nativ

eA

mer.

——

�0.3

78

0.5

33

0.4

79

�0.1

58

0.5

41

0.7

71

Asi

an

——

0.6

48

0.4

20

0.1

23

0.4

75

0.4

27

0.2

66

Afr

ican

Am

eric

an

——

0.8

20n

0.0

96

0.0

00

0.7

59n

0.0

98

0.0

00

His

panic

——

0.6

07n

0.2

35

0.0

10

0.5

88n

0.2

39

0.0

14

Contr

olva

riab

les

No

Yes

No

Yes

N8,7

45

8,7

45

8,7

45

8,7

45

Log–l

ikelih

ood

�5743.0

3,

df5

5,

p5

0.0

00

�5580.2

3,

df5

15,

p5

0.0

00

�5704.0

0,

df5

9,

p5

0.0

00

�5549.0

9,

d,

p5

0.0

00f5

19

NO

TE:

Mod

el1

:C

ox

and

Snell

R2

50.0

70.

Nagelk

erk

eR

25

0.0

92,

McFadd

ens

R2

50.0

52.

McK

elv

ey

and

Zavo

ina’s

R2

50.0

84.

Efron’s

R2

50.0

71.

Mod

el2:

Cox

and

Snell

R2

50.1

03.

Nagelk

erk

eR

25

0.1

38.

McFad

dens

R2

50.0

79.

McK

elv

ey

and

Zavo

ina’s

R2

50.1

29.

Efron’s

R2

50.1

07.

Model3:

Cox

and

Snell

R2

50.0

77.

Nagelk

erk

eR

25

0.1

03.

McFadd

ens

R2

50.0

57.

McK

elv

ey

and

Zavo

ina’s

R2

50.0

95.

Efron’s

R2

50.0

79.

Mod

el

4:

Cox

and

Snell

R2

50.1

10.

Nagelk

erk

eR

25

0.1

46.

McFadd

ens

R2

50.0

84.

McK

elv

ey

and

Zavo

ina’s

R2

50.1

37.

Efron’s

R2

50.1

13.

For

mod

els

that

inclu

ded

the

contr

olva

riable

s,th

ed

ecis

ion

was

mad

eto

illust

rate

only

the

est

imate

sw

ithth

etr

ansf

orm

ed

varia

ble

sbase

don

the

resu

ltsofa

BIC

test

.The

fitst

at

com

mand

inS

TATA

was

use

dto

com

pare

the

modela

fter

ad

just

ing

for

nonlin

earit

ies

on

the

rightsi

de

ofth

eeq

uatio

n(L

ong

and

Fre

ese

,2006).

The

resu

ltsin

dic

ate

dth

atth

ed

iffere

nce

in12.9

22

inB

ICpro

vided

very

stro

ng

sup

port

for

the

modelw

ithth

etr

ansf

orm

ed

varia

ble

s.The

fitte

stw

as

als

ouse

dto

com

pare

the

fitof

the

models

with

the

contr

ol

varia

ble

s,w

ith(M

odel2)

and

with

out

(Mod

el4)

inte

ractio

neffe

cts

.The

diff

ere

nce

of

25.9

82

inB

ICpro

vided

stro

ng

sup

port

for

Mod

el4.

The

ove

rall

perc

ent

corr

ectly

pre

dic

ted

by

Model4

was

66.4

perc

ent,

sensi

tivity

571.5

perc

ent,

specifi

city

561

perc

ent,

fals

ep

osi

tive

533.7

perc

ent,

fals

enegativ

e5

33.4

perc

ent.

796 Social Science Quarterly

Page 16: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

TA

BLE

2

Tra

nsfo

rmation

ofth

eE

ffects

ofR

ace,E

thnic

ity,and

CR

AA

ctivity

on

Apply

ing

toa

Regula

ted

Lender,

Odds

Ratio

(OR

),P

erc

entC

hange

(%C

hg)

inO

dds,

and

Confidence

Inte

rvals

(CI)

Racia

land

Eth

nic

Chara

cte

ristic

s

Mod

el1

Mod

el2

Mod

el3

Mod

el4

OR

%C

hg

C.I.

OR

%C

hg

C.I.

OR

%C

hg

C.I.

OR

%C

hg

C.I.

Am

er.

Ind

.,A

lask

an,

Pac.

Isl.,

Haw

aiia

n,

or

Nativ

eA

mer.

0.4

92

n�

50.8

0.3

00,

0.5

42n�

45.8

0.3

27,

0.6

70�

33.0

0.2

93,

0.6

38�

36.2

0.2

75,

0.8

07

0.8

97

1.5

34

Asi

an

2.0

18n

101.8

1.4

03,

2.1

75n

117.5

1.5

00,

1.6

00n

60.0

1.0

46,

1.8

06

n80.6

1.1

70,

2.9

03

3.1

52

2.4

50

2.7

87

Afr

ican

Am

eric

an

0.3

49n�

65.1

0.3

18,

0.5

90n�

41.0

0.5

24,

0.2

30n�

77.0

0.2

01,

0.3

96

n�

60.4

0.3

38,

0.3

83

0.6

65

0.2

64

0.4

64

His

panic

1.1

98

19.8

0.9

72,

1.5

33n

53.3

1.2

35,

0.8

51�

14.9

0.5

78,

1.0

83

8.3

0.7

29,

1.4

76

1.9

03

1.2

53

1.6

09

CR

Aactiv

ity0.8

75n�

12.5

0.8

01,

0.9

76

�2.4

0.8

80,

0.6

23n�

37.7

0.5

54,

0.7

00

n�

30.0

0.6

12,

0.9

56

1.0

83

0.7

02

3.7

15

Inte

ractio

ns:

CR

AA

ctiv

ityand

Race/E

thnic

ity

Am

er.

Ind

.,A

lask

an,

Pac.

Isl.,

Haw

aiia

n,

or

Nativ

eA

mer.

——

0.6

86�

31.5

0.2

41,

0.8

54�

14.6

0.2

96,

1.9

50

2.4

66

Asi

an

——

1.9

12

91.2

0.8

40,

1.6

09

60.9

0.6

97,

4.3

54

3.7

15

Afr

ican

Am

eric

an

——

2.2

70n

127.0

1.8

82,

2.1

37

n113.7

1.7

62,

2.7

37

2.5

92

His

panic

——

1.8

35n

83.5

1.1

58,

1.8

00

n79.9

1.1

26,

2.9

08

2.8

77

Contr

olva

riab

les

No

Yes

No

Yes

N8,7

45

8,7

45

8,7

45

nIn

dic

ate

sa

statis

tically

signifi

cant

varia

ble

at

the

5p

erc

ent

leve

l.

Racial Disparities in Mortgage Credit 797

Page 17: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

Overall Models

As reported in Table 1, overall, all models were significant at the 0.001level. Based on the model summary statistics, the models that included thecontrol variables, Models 2 and 4, were preferred. The results of two separatelikelihood ratio tests on the key independent variables of interest suggest theeffect of an applicant’s race and ethnicity is significant at the 0.001 level(LRX2 5 199.97, df 5 8, po0.001), and the effect of CRA activity is sig-nificant at the 0.001 level (LRX2 5 62.49, df 5 5, po0.001), suggesting thatboth of the key variables of interest should exert a significant effect on the typeof lender to which an applicant applies. In Model 2, which considers race andethnicity and CRA activity as additive factors, an applicant’s race and ethnicitydoes appear to influence the likelihood of applying to a regulated lender;however, CRA activity is only marginally significant. When allowing CRAactivity to interact with an applicant’s race or ethnicity as in Model 4, CRAactivity significantly influenced the effects of race and ethnicity. To determinewhich of these models is preferred, model summary statistics and the resultsfrom a STATA fittest were compared. The fittest results suggest that Model 4,significant at the 0.001 level (LRX2 5 58.86, df 5 4, po0.001), is the pre-ferred model. Overall summary statistics for Model 4 are presented in Table 1.

Interpretation of Effects

Table 2 presents the transformation of the parameter estimates as thepercent change in the odds and the odds ratio for all four models; however,

TABLE 3

Change in the Predicted Probabilities of Applying to a Regulated Lender by Race,Ethnicity, and CRA Activity

Controlling for Applicant CharacteristicsPr (y 5 1|x)

C. Change,95% CI for ChangeApplicant’s Racial and Ethnic Affiliation

A. CRAActivity

B. No CRAActivity

White, non-Hispanic 0.511 0.600 � 0.088n

[� 0.121, � 0.055]African American 0.470 0.372 0.098 n

[0.061, 0.135]Hispanic 0.671 0.618 0.053

[� 0.054, 0.159]Asian 0.752 0.730 0.023

[� 0.135, 0.180]Amer. Ind., Alaskan, Pac. Isl., Hawaiian,

or Native Amer.0.363 0.488 � 0.125

[� 0.382, 0.132]

nIndicates a statistically significant variable based on the range of the confidence interval.

798 Social Science Quarterly

Page 18: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

TA

BLE

4

Eff

ects

of

Contr

olV

ariable

son

Apply

ing

toa

Regula

ted

Lender

Contr

olVaria

ble

s

Mod

el2

Mod

el4

Coef.

s.e.

pO

R|%

Chg

Coef.

s.e.

pO

R|%

Chg

Pop

ula

tion

�0.3

06

n0.0

61

0.0

00

0.7

36

�0.3

15

n0.0

62

0.0

00

0.7

30

�26.4

�27.0

Ap

plic

ant

sex

0.0

96

n0.0

47

0.0

40

1.1

01

0.0

99

n0.0

47

0.0

36

1.1

04

10.1

10.3

%C

olle

ge

ed

.in

Censu

str

act

0.0

47

n0.0

07

0.0

00

1.0

48

0.0

45

n0.0

08

0.0

00

1.0

46

4.8

4.6

%C

olle

ge

ed

.tr

ansf

orm

ed

�0.0

01

n0.0

00

0.0

00

0.9

99

�0.0

01

n0.0

08

0.0

00

0.9

99

�0.1

0�

0.1

0Tr

act

incom

era

tio0.0

07

n0.0

02

0.0

00

1.0

07

0.0

08

n0.0

02

0.0

00

1.0

08

0.7

00.8

0%

Min

orit

yp

op

.in

Censu

str

act

0.0

06

0.0

04

0.1

45

1.0

06

0.0

07

0.0

04

0.0

68

1.0

07

0.6

00.7

0%

Min

orit

ytr

ansf

orm

ed

�0.0

00

n0.0

00

0.0

04

1.0

00

�0.0

00

n0.0

00

0.0

01

1.0

00

00

Ap

plic

ant

incom

e0.0

02

n0.0

01

0.0

07

1.0

02

0.0

02

n0.0

01

0.0

12

1.0

02

0.2

00.2

0#

Housi

ng

units

inC

ensu

str

act

0.0

01

n0.0

00

0.0

01

1.0

01

0.0

01

n0.0

00

0.0

00

1.0

01

0.1

00.1

0#

Housi

ng

units

transf

orm

ed

�0.0

00

0.0

00

0.4

70

1.0

00

�0.0

00

0.0

00

0.2

40

1.0

00

00

Const

ant

�0.7

80

n0.2

14

0.0

00

—�

0.6

78

n0.2

15

0.0

02

—N

8,7

45

8,7

45

nIn

dic

ate

sa

statis

tically

signifi

cant

varia

ble

at

the

5p

erc

ent

leve

l.

Racial Disparities in Mortgage Credit 799

Page 19: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

following Long and Freese (2006), we focus our interpretation only on thepercent change in odds. The vertical shaded boxes illuminate the percentchange in odds for each of the models. The negative sign suggests the oddsof applying to a regulated lender decrease based on the value of the in-dependent variable and, conversely, when the sign is positive, it suggests theodds of applying to a regulated lender increase.

When considering the interaction between race/ethnicity and CRA ac-tivity (Model 4), for African-American applicants, the odds of applying to aregulated lender increase by 113.7 percent, when controlling for applicantcharacteristics and holding all other variables constant, compared to the basegroup. This stands in stark contrast to the percent change in Model 2, whichdoes not consider the interactive effect of CRA activity, and suggests that theodds of African Americans applying to a regulated lender decrease by 41.0percent. For Hispanic applicants in a city with CRA activity, the odds ofapplying to a regulated lender increase by 79.9 percent. The odds of Asianand American-Indian or Alaskan-Native, Pacific-Islander or Hawaiian, orNative-American applicants applying to a regulated lender are not signif-icantly influenced by CRA activity. As evidenced by the difference betweenModels 3 and 4, applicant characteristics do influence the probability ofapplying to a regulated lender; however, the effects of race and ethnicityand CRA activity are still significant for African-American and Hispanicapplicants.

When interpreting models with interaction effects, it is also important toconsider the magnitude of the individual variables of interest. As reported inTable 2, the percent change in odds for CRA activity suggests that thelikelihood of applying to a regulated lender decreases in a city with CRAactivity by 30 percent; however, the magnitude and direction of these effectsvary based on an applicant’s race or ethnicity. To better understand thecombined effect of race/ethnicity and CRA activity, predicted probabilities,change in probabilities, and confidence intervals were computed usingModel 4 and are reported in Table 3. Using the prvalue command inSTATA, we computed predicted probabilities for each racial and ethnicgroup when CRA activity was 1 or 0. The confidence intervals are providedto determine whether the effects are significant when accounting for theinteraction. All other variables were held at their mean values. For African-American and white applicants, CRA activity has a significant effect, as thesegroups experience the largest difference in the discrete change, as reported inColumn C. The predicted probabilities reported in Column A suggest thatwhen CRA activity is present, the probability that African Americans applyto a regulated lender approaches the probability of white applicants, butremains slightly less (0.47 compared to 0.51, respectively). For African-American applicants, CRA activity increases the probability of applying to aregulated lender by 0.10, and is significant based on a 95 percent confidenceinterval. Conversely, for white, non-Hispanic applicants, applying for a loanin an area with CRA activity decreases the probability of applying to a

800 Social Science Quarterly

Page 20: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

regulated lender by 0.09, again significant based on a 95 percent confidenceinterval. The gap between predicted probabilities of applying to a regulatedlender is much greater among these groups when CRA activity is not pres-ent, as presented in Column B. For Asian and American Indians or AlaskanNatives, Pacific Islanders or Hawaiians, or Native Americans, CRA activitydoes not have a significant effect on the probability of applying to a reg-ulated lender, above and beyond the effects of race. For Hispanic borrowers,while the predicted probability suggests an increased probability of applyingto a regulated lender, we can be less confident about the predicted prob-ability as the confidence interval crosses the threshold of 0, and the valuesreported in Columns A and B only slightly differ.

The effects of the control variables were in the anticipated direction andare presented in Table 4. Controlling for demographic characteristics andother characteristics that influence applicant behavior did not eliminate theeffects of race and CRA activity.

Discussion and Conclusion

In this article, we considered the local context in which borrowers pursuelending options—focusing on the effects of community reinvestment ac-tivity. Overall, we agree with Dymski (1999) that CRA is only part of thepuzzle and that the lending environment is indeed complex. Our findingsidentify an additional factor in that complexity: the local actors that in-fluence community reinvestment efforts. Our findings also support previouswork by Williams, McConnell, and Nesiba (2001) on the potential of CRAactivism and organizing to influence differential outcomes. In a city withcommunity reinvestment activity, African-American applicants are morelikely to apply to CRA-regulated lenders than their counterparts in a citywith low community reinvestment activity, even when controlling for ap-plicant characteristics. For Asian, Hispanic, and applicants identifying asAmerican Indian, Alaskan Native, Pacific Islander, Hawaiian, or NativeAmerican, CRA activity does not exert a significant effect. However, ourfindings also suggest CRA activity may not be a panacea. Even in placeswhere reinvestment activity is present and African-American applicantsmake gains, they are still less likely than white applicants to pursue creditfrom CRA-regulated lenders.

The comparison of the effects across racial and ethnic groups is partic-ularly interesting. In cities with CRA activity, the probability that whiteapplicants pursue access to credit from regulated lenders declines. Scholarsof racism have argued that racial oppression ultimately adversely affectsdisadvantaged people of all ethnic groups. Feagin, Vera, and Batur (2001)argue that everyone in society suffers from racism because it prevents raciallyoppressed people from realizing their full potential. Therefore, racism resultsin what they call ‘‘societal waste,’’ the loss of human talent and contributions

Racial Disparities in Mortgage Credit 801

Page 21: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

to society that would benefit everyone. However, our finding implies some-thing a bit different. It is not that whites may face a cost because of societalracism but because of efforts to combat it. Thus, our finding is more in linewith Stephen Steinberg’s (2001) analysis of affirmative action in highereducation. He argues:

when members of racial minorities were excluded, it was because they werestigmatized as inferior or undesirable, and their exclusion was part of asystemic pattern of exploitation and domination. No such claims can bemade with respect to the small number of whites adversely affected byaffirmative action . . . All that can be said is that, as individuals, they paidthe price of a societal commitment to achieving racial parity . . . and whilesome whites are injured in the process, it is due not to racism, but to aneffort to combat racism. (2001:252)

Based on our approach in this article, our results suggest that CRA activitymay help connect African-American applicants to regulated lenders; how-ever, this may come at some cost to white applicants. However, we arelimited to understanding the true racialized costs of applying to a nonreg-ulated versus a regulated lender. First, we face data limitations in that wecannot precisely determine to what extent our results are driven by the factthat members of racial and ethnic groups apply to regulated lenders, but arein turn denied, efforts that may lead them to seek credit from nonregulatedlenders. Given the similar loan origination rates by regulated lenders in bothcities, among all applicants, it does not appear that systematic biases exist atthe city level. However, when considering the percentage of African-Amer-ican applications that were approved, regulated lenders in Cleveland orig-inated a greater percentage of loans for African-American applicants (80percent) than regulated lenders in St. Louis (72 percent). This discrepancymight suggest that African-American applicants in St. Louis may be drivento the nonregulated marketplace at a higher rate than those in Cleveland,which further supports our argument that CRA activity, or lack thereof,exerts an influence, but we are not able to empirically capture these effects.Second, it is possible that whites applying to nonregulated lenders receivefairer deals and far better treatment than African Americans applying to thesame lenders. Thus, the effect of CRA activity may not be as simple as azero-sum game, where gains for people of color come with clear losses forwhites, and is one that warrants further exploration.

It is important to digest these findings as they relate to the larger pictureof the overall relationship between race, ethnicity, and access to main-stream financial services. Although the probability that white applicantspursue access to credit from regulated lenders declines in light of CRAactivity, the probability is still slightly higher than that of African-Amer-ican applicants. Even with CRA activity, while this gap narrows, it has notbeen eliminated. Given that regulated lenders are more likely to originateprime loans, white applicants are still more likely to accrue the benefits of

802 Social Science Quarterly

Page 22: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

being served by traditional lenders. This finding suggests that althoughCRA activity is present, in the current financial services environment, thereare still limits to what this activity can do to create a more equitablelending environment.

Finally, undoubtedly, there remain some challenges to discerning thespecifics of these racialized effects, particularly as they relate to what we caninfer from the construction of racial and ethnic categories. We had an-ticipated that Hispanic applicants might also be negatively impacted by alack of CRA activity. The results did not indicate this, and we suspect thisfinding may be due to the limitations that exist in understanding theunique characteristics of the racial groups identifying as Hispanic. Forexample, we cannot discern from this analysis to what extent the physicalracialized characteristics of applicants identifying as Hispanic may affecttheir access to credit, or what particular subgroups of this category aredriving the results. Future research that breaks this category down furthermay find that different subgroups of Hispanics have different outcomes.Likewise, while the analysis illustrates that Asians are more likely thanwhites to apply to a regulated lender, the category of Asian is also prob-lematic for similar reasons. This category combines several different peo-ples with very different lived experiences and few studies break the categorydown sufficiently. This often perpetuates the model minority myth, whichjuxtaposes people of color against each other (see Chou and Feagin, 2008;Kerbo, 2008; Steinberg, 2001). Based on our findings and conclusion,there is a need for future research that focuses on the specific effects of theactors engaged in CRA efforts. Likewise, a greater distinction is neededbetween the interests involved and the racial and ethnic groups they rep-resent. It could very well be that the effects for African-American bor-rowers in a city such as Cleveland are due to the involvement of groupsrepresenting the interests of African Americans and, as such, CRA activityin other places may lead to different outcomes if the interests representeddiffer.

Dymski and Aldana (2007) argue that minorities in cities with largerpopulations of minority residents might gain some market power advan-tages, but this comparison of two cities with similar racial and ethnic char-acteristics and populations suggests that population proportion alone maynot be enough; rather, organizing and activism around community rein-vestment may yield certain advantages to groups. Thus, it is important tounderstand how the community reinvestment environment shapes the op-portunities that become available to minority groups from both lendersregulated under the guise of CRA and those not subject to CRA. Second,additional research needs to explore the private actors that compose effortsand their variation across places. In the case of the latter, does this lead tosome advantages or disadvantages for certain racial or ethnic groups basedon the interests or mission of the organizations that shape the environment?These are questions that remain to be explored.

Racial Disparities in Mortgage Credit 803

Page 23: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

REFERENCES

Barr, M. 2004. ‘‘Banking the Poor.’’ Yale Journal on Regulation 21(1):121–237.

Benston, G. 1997. ‘‘Discrimination in Mortgage Lending: Why HMDA and CRA Should beRepealed.’’ Journal of Retail Banking Services 19(3):47–57.

Bostic, R., and B. Robinson. 2005. ‘‘What Makes Community Reinvestment Act AgreementsWork?’’ Housing Policy Debate 16(3 and 4):513–45.

Bovaird, T. 2007. ‘‘Beyond Engagement and Participation: User and Community Copro-duction of Public Services.’’ Public Administration Review 67(5):846–60.

Bright, E. 2003. Reviving America’s Forgotten Cities. London: Routledge.

Bunce, H., D. Bruenstein, C. Herbert, and R. Scheessele. 2001. Subprime Foreclosures:The Smoking Gun of Predatory Lending? Washington, DC: U.S. Department of Housingand Urban Development. Available at hhttp://www.huduser.org/publications/pdf/brd/12Bunce.pdfi.Bureau of Labor Statistics. 2008. Unemployment Rates for the Largest 50 Cities: Annual AverageRankings, Year: 2003. Washington, DC: Bureau of Labor Statistics. Available at hhttp://www.bls.gov/lau/lacilg03.htmi.Casey, C. 2009. ‘‘Linking Social Capital and Indirect Policy Tools: Fostering EquitableCommunity Development Responses?’’ Journal of Planning Education and Research 28:413–25.

Chou, R., and J. Feagin. 2008. The Myth of the Model Minority: Asian Americans FacingRacism. Boulder, CO: Paradigm Publishers.

Courchane, M., B. Surette, and P. Zorn. 2004. ‘‘Subprime Borrowers: Transition and Out-comes.’’ Journal of Real Estate Finance and Economics 29(4):365–92.

Dymski, G. 1999. The Bank Merger Wave: The Economic Causes and Social Consequences ofFinancial Consolidation. Armonk, NY: M.E. Sharpe.

Dymski, G., and C. Aldana. 2007. ‘‘The Racial-U Curve in U.S. Residential Credit Marketsin the 1990s: Empirical Evidence from a Post Keynesian World.’’ Pp. 70–99 in R. Holt andS. Pressman, eds., Empirical Post Keynesian Economics: Looking at the Real World. Armonk,NY: M.E. Sharpe.

Feagin, J. R., H. Vera, and P. Batur. 2001. White Racism: The Basics. New York: Routledge.

Flippen, C. 2001. ‘‘Racial and Ethnic Inequality in Homeownership and Housing Equity.’’Sociological Quarterly 42(2):121–49.

Gruenstein, D., and C. Herbert. 2000. Analyzing Trends in Subprime Originations andForeclosures: A Case Study of the Atlanta Metro Area. Washington, DC: Abt Associates.

Gunningham, N., R. Kagan, and D. Thornton. 2004. ‘‘Social License and EnvironmentalProtection: Why Businesses Go Beyond Compliance.’’ Law and Social Inquiry 29:307–27.

Haines, F., N. Reichman, and C. Scott. 2008. ‘‘Problematizing Authority and Legitimacy inLaw and Policy.’’ Law and Policy 30(1):1–11.

Holyoke, T. 2004. ‘‘Community Mobilization and Credit: The Impact of Nonprofits andSocial Capital on Community Reinvestment Act Lending.’’ Social Science Quarterly 85:187–205.

Hula, R., and C. Jackson-Elmoore. 2001. ‘‘Governing Nonprofits and Local Political Pro-cesses.’’ Urban Affairs Review 36(3):324–58.

804 Social Science Quarterly

Page 24: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

Immergluck, D. 2004. Credit to the Community: Community Reinvestment and Fair LendingPolicy in the United States. Armonk, NY: M.E. Sharpe, Inc.

Joint Center for Housing Studies. 2002. The 25th Anniversary of the Community ReinvestmentAct: Access to Capital in an Evolving Financial Services System. Cambridge, MA: Joint Centerfor Housing Studies. Available at hhttp://www.jchs.harvard.edu/publications/government-programs/cra02-1.pdfi.Kerbo, H. 2008. Social Stratification and Inequality, 7th ed. Columbus, OH: McGraw Hill.

Lang, W., and L. Nakamura. 1993. ‘‘A Model of Redlining.’’ Journal of Urban Economics33:223–34.

Long, J., and J. Freese. 2006. Regression Models for Categorical Dependent Variables UsingStata, 2nd ed. College Station, TX: Stata Press.

Marsico, R. 2005. Democratizing Capital: The History, Law and Reform of the CommunityReinvestment Act. Durnham, NC: Carolina Academic Press.

National Community Reinvestment Coalition (NCRC). 2005. CRA Commitments. Wash-ington, DC: NCRC.

———. (2007). Income is No Shield: Against Racial Differences in Lending: A Comparison ofHigh-Cost Lending in America’s Metropolitan Areas. Available at hhttp://www.ncrc.orgi.Nichols, J., A. Pennington-Cross, and A. Yezer. 2004. ‘‘Borrower Self-Selection, Under-writing Costs, and Subprime Mortgage Credit Supply.’’ Journal of Real Estate Finance andEconomics 30(2):197–291.

Norton, E., H. Wang, and C. Ai. 2004. ‘‘Computing Interaction Effects and Standard Errorsin Logit and Probit Models.’’ Stata Journal 4:154–67.

Orr, M. 2001. ‘‘BUILD: Governing Nonprofits and Relational Power.’’ Policy Studies Review18(4):71–90.

Salamon, L., ed. 2002. The Tools of Government. New York: Oxford Press.

Schneider, A., and H. Ingram. 1997. Policy Design for Democracy. Lawrence, KS: Universityof Kansas Press.

Schwartz, A. 1998. ‘‘Banking Lending to Minority and Low-Income Households andNeighborhoods: Do Community Reinvestment Agreements Make a Difference?’’ Journal ofUrban Affairs 20(3):269–301.

———. 2000. ‘‘The Past and Future of Community Reinvestment Agreements.’’ HousingFacts and Findings 2(1):3–7.

Shadish, W. R., T. D. Cook, and D. T. Campbell. 2002. Experimental and Quasi-Exper-imental Designs for Generalized Causal Inference. Boston, MA: Houghton-Mifflin.

Shlay, A. 1999. ‘‘Influencing the Agents of Urban Structure: Evaluating the Effects ofCommunity Reinvestment Organizing on Bank Residential Lending Practices.’’ Urban Af-fairs Review 35(2):247–78.

Squires, G., ed. 1992. From Redlining to Reinvestment: Community Responses to Urban Dis-investment. Philadelphia, PA: Temple University Press.

———, ed. 2003. Organizing Access to Capital: Advocacy and Democratization of FinancialInstitutions. Philadelphia, PA: Temple University Press.

Steinberg, S. 2001. The Ethnic Myth: Race, Ethnicity, and Class in America. Boston, MA:Beacon Press.

Racial Disparities in Mortgage Credit 805

Page 25: Racial Disparities in Access to Mortgage Credit: Does Governance Matter?

Stoker, G. 1998. ‘‘Governance as Theory: Five Propositions.’’ International Social ScienceJournal 155:17–28.

Temkin, K., J. Johnson, and D. Levy. 2002. Subprime Markets, the Role of GSE’s, and Risk-Based Pricing. Washington, DC: Urban Institute and the U.S. Department of Housing andUrban Development, Office of Policy Development and Research.

U.S. Department of Housing and Urban Development. 2000. Unequal Burden: Income andRacial Disparities in Subprime Lending in America. Available at hhttp://www.huduser.org/publications/fairhsg/unequal.htmli.Williams, R. 2009. ‘‘Using Heterogenous Choice Models to Compare Logit and ProbitCoefficients Across Groups.’’ Sociological Methods & Research 37:531–59.

Williams, R., D. E. McConnell, and R. Nesiba. 2001. ‘‘The Effects of the GSEs, CRA, andInstitutional Characteristics on Home Mortgage Lending to Underserved Markets.’’ City-scape: A Journal of Policy Development and Research 5(3):9–106.

Williams, R., and R. Nesiba. 1997. ‘‘Racial, Economic, and Institutional Differences inHome Mortgage Loans: St. Joseph County Indiana.’’ Journal of Urban Affairs 19(1):73–103.

Williams, R., R. Nesiba, and D. E. McConnell. 2005. ‘‘The Changing Face of Inequality inHome Mortgage Lending.’’ Social Problems 52(2):181–208.

Wyly, E., M. Atia, H. Foxcroft, D. Hammel, and K. Phillips-Watts. 2006. ‘‘AmericanHome: Predatory Mortgage Capital and Neighbourhood Spaces of Race and Class Exploi-tation in the United States.’’ Geografjska. Annaler 88B(1):105–13.

806 Social Science Quarterly